RAG context can surface memorized training continuations hidden in ordinary testing

Evidence record

Primary or indexed evidence for Radar finding DB-RADAR-CP2-CP3-076. Controlled open-weight models; no demonstrated production personal-data breach.

Read the original source ↗

Claim-level citations (3)

• supportsRetrieved context changes which memorized continuations become extractable: The study finds both context-robust and context-sensitive memorization when evaluating fixed prefix-suffix pairs with retrieved material.

  Abstract, paired item-level probabilistic suffix extraction

• supportsRetrieved context changes which memorized continuations become extractable: Retrieved contexts in the experiment suppressed some marginal extraction cases but also enabled new extractable continuations relative to prefix-only tests.

  Abstract, context-sensitive boundary findings

• supportsRetrieved context changes which memorized continuations become extractable: The experiments score fixed training-text continuations probabilistically across three open-weight instruction-tuned models, so extractability in the paper does not itself establish emitted production leakage.

  Abstract, paired prefix-suffix scoring method